Engaged governance participants were identified as the top 25% of governance voters ranked by voting frequency. Several onchain actions set engaged governance participants apart from the rest, most importantly voting frequency, number of votes, and number of delegates pledged to the participant. Other indicators of engaged governance participants include ENS registrations and the number of decentralized exchange (DEX) trades and NFT activity within the last 60 days.

This research focuses on voting behaviors within the Optimism Token House, part of the Optimism Collective, which uses token-weighted voting for governance. To measure voter engagement, a single value representing voting frequency was calculated.  This calculation compares a participant’s number of votes to the total votes cast each month to find their share of total votes. The share of votes is calculated for each month and averaged across all months to provide a single engagement score. This allows fair comparison of participant voter patterns without unfairly penalizing newer or older participants.  Other analyzed variables relating to voting behavior were number of votes and number of delegates pledged.

Onchain actions were identified as having an ENS name, making a proposal, making votes in other DAOs, donations through Gitcoin, and NFT activity and DEX trades within the last 60 days.

Data collection started on Dune with initial SQL queries to obtain a list of voters and identify those voter addresses in the datasets of other onchain actions (DEX trades, proposals, etc).  This data was retrieved via API by a Python script for statistical analysis and logistic regression modeling.

The correlation analysis and logistic regression metrics discussed in this research are based on the data and models implemented in the GitHub repository: https://github.com/BrandynHamilton/superchain_research

Correlation values show the relationship between two variables. Closer to 1 is a strong positive correlation, closer to -1 is a strong negative correlation, and close to 0 means no correlation. Weak correlation values were observed among the selected onchain actions. Focusing on indicators of engaged participants, the strongest correlation for voting frequency was the number of votes a participant has made, with a value of 0.15.  The strongest correlation for number of votes was the participant’s voting frequency, with a value of 0.15.  The strongest correlation for the number of proposals was the number of pledged delegates to the voter, with a value of 0.23.

Logistic regression predicts the probability of a binary outcome based on input features. Our model, trained to identify engaged participants (top 25% by voting frequency), achieved a ROC AUC score of 0.77. This means there is a 77% chance that a randomly selected engaged participant is ranked higher in probability than a non-engaged one. The recall score of 0.58 indicates that the model correctly identifies 58% of actual engaged participants.

This research has identified the key behaviors and on-chain actions that set apart highly engaged governance participants in the Optimism Token House. Our findings show that consistent voting and active involvement in governance activities are crucial indicators of engagement. Notably, voting frequency, number of votes, and delegates pledged to a participant emerged as the most significant predictors of engagement. Additionally, other on-chain actions, such as decentralized exchange trades and NFT activities, also played a role in identifying engaged participants.

Using logistic regression, we gained valuable insights into predicting engagement, highlighting the diverse range of on-chain actions that influence governance participation. The model’s 77% ROC AUC score indicates a strong capability to differentiate between engaged and non-engaged participants, reflecting the complexity and multi-dimensional nature of governance behavior. Moving forward, we could further improve predictive accuracy by exploring more data sources and refining our engagement metrics to better capture the full spectrum of voter activity.

Github: https://github.com/BrandynHamilton/superchain_research
Dune Dashboard: https://dune.com/brandyn/optimism-voter-analysis

Engaged governance participants were identified as the top 25% of governance voters ranked by voting frequency. Several onchain actions set engaged governance participants apart from the rest, most importantly voting frequency, number of votes, and number of delegates pledged to the participant. Other indicators of engaged governance participants include ENS registrations and the number of decentralized exchange (DEX) trades and NFT activity within the last 60 days.

This research focuses on voting behaviors within the Optimism Token House, part of the Optimism Collective, which uses token-weighted voting for governance. To measure voter engagement, a single value representing voting frequency was calculated.  This calculation compares a participant’s number of votes to the total votes cast each month to find their share of total votes. The share of votes is calculated for each month and averaged across all months to provide a single engagement score. This allows fair comparison of participant voter patterns without unfairly penalizing newer or older participants.  Other analyzed variables relating to voting behavior were number of votes and number of delegates pledged.

Onchain actions were identified as having an ENS name, making a proposal, making votes in other DAOs, donations through Gitcoin, and NFT activity and DEX trades within the last 60 days.

Data collection started on Dune with initial SQL queries to obtain a list of voters and identify those voter addresses in the datasets of other onchain actions (DEX trades, proposals, etc).  This data was retrieved via API by a Python script for statistical analysis and logistic regression modeling.

The correlation analysis and logistic regression metrics discussed in this research are based on the data and models implemented in the GitHub repository: https://github.com/BrandynHamilton/superchain_research

Correlation values show the relationship between two variables. Closer to 1 is a strong positive correlation, closer to -1 is a strong negative correlation, and close to 0 means no correlation. Weak correlation values were observed among the selected onchain actions. Focusing on indicators of engaged participants, the strongest correlation for voting frequency was the number of votes a participant has made, with a value of 0.15.  The strongest correlation for number of votes was the participant’s voting frequency, with a value of 0.15.  The strongest correlation for the number of proposals was the number of pledged delegates to the voter, with a value of 0.23.

Logistic regression predicts the probability of a binary outcome based on input features. Our model, trained to identify engaged participants (top 25% by voting frequency), achieved a ROC AUC score of 0.77. This means there is a 77% chance that a randomly selected engaged participant is ranked higher in probability than a non-engaged one. The recall score of 0.58 indicates that the model correctly identifies 58% of actual engaged participants.

This research has identified the key behaviors and on-chain actions that set apart highly engaged governance participants in the Optimism Token House. Our findings show that consistent voting and active involvement in governance activities are crucial indicators of engagement. Notably, voting frequency, number of votes, and delegates pledged to a participant emerged as the most significant predictors of engagement. Additionally, other on-chain actions, such as decentralized exchange trades and NFT activities, also played a role in identifying engaged participants.

Using logistic regression, we gained valuable insights into predicting engagement, highlighting the diverse range of on-chain actions that influence governance participation. The model’s 77% ROC AUC score indicates a strong capability to differentiate between engaged and non-engaged participants, reflecting the complexity and multi-dimensional nature of governance behavior. Moving forward, we could further improve predictive accuracy by exploring more data sources and refining our engagement metrics to better capture the full spectrum of voter activity.

Github: https://github.com/BrandynHamilton/superchain_research
Dune Dashboard: https://dune.com/brandyn/optimism-voter-analysis